apply-job

name	apply-job
description	Generate tailored CV and cover letter for job applications. Use when user says /apply, asks to write a cover letter, wants to apply to a job posting, needs help with job applications, or provides a job URL. Orchestrates job analysis, company research, and document generation following Swiss formal style rules.
degree-of-freedom	low
allowed-tools	Read, Write, Edit, mcp__mcp-server-firecrawl__*

@.claude/skills/apply-job/references/job-types.md

Apply Job Skill

Generate tailored, human-like cover letters and CV recommendations for job applications.

---

How This Skill Works

This skill uses lazy loading to minimize context usage. Context files are loaded only when needed.

Workflow:

1. Analyze job & classify → 2. Clarify with user → 3. Research company → 4. Generate content → 5. Validate quality → 6. Create output

Context Loading Points:

• Startup: job-types.md only (~800 tokens)
• After classification: profile-essential.md (~2,000 tokens)
• After clarification: writing-rules.md (~1,500 tokens)
• If Swiss company: swiss-market-rules.md (~600 tokens)

---

Skill Interface

Invocation: /apply [job-url-or-description]

Inputs:

• Job URL (scraped via Firecrawl MCP)
• OR pasted job description
• User responses to clarification questions

Outputs (Generated by Claude - ALL MANDATORY):

• job-description.md - Scraped job posting
• cover-letter.md - Tailored cover letter
• cover-letter.html - HTML version
• cover-letter.pdf - MANDATORY PDF (max 2 pages, print-ready)
• cv-{variant}.html - MANDATORY Copy of CV variant
• cv-{variant}.pdf - MANDATORY CV PDF (max 2 pages)
• application-email.md - Application email draft

---

Step 1: Analyze Job & Classify

Get the job description:

• If URL provided: Scrape with Firecrawl MCP
• If pasted: Parse key requirements

Classify job type by scanning for keywords (see job-types.md for details):

Type	Keywords	CV Variant
A: Digital Health	health, medical, wearable, clinical	cv-simon-yang-digital-health.html
B: AI/ML	AI, ML, data science, analytics	cv-simon-yang-ai-product.html
C: GenAI/DevTools	GenAI, LLM, developer, automation	cv-simon-yang-genai-tools.html
D: ESG/Climate	ESG, climate, risk, sustainability	cv-simon-yang-reprisk.html

Gate 1 Check:

• Job type classified with confidence >70%
• CV variant selected
• ≥3 matching keywords identified

After Gate 1 passes, load: references/profile-essential.md

---

Step 2: Clarify with User

Ask these questions before generating:

1. "I've classified this as [Type X]. Is that correct?"
2. "What 2-3 aspects of this role interest you most?"
3. "Any specific achievement you want to emphasize?"
4. "Company tone: startup casual, corporate formal, or Swiss formal?"
5. "Any gaps or concerns to address proactively?"

Wait for user responses before proceeding.

After clarification, load: references/writing-rules.md

---

Step 3: Research Company

Use Firecrawl MCP to scrape:

• Company website (About, Products, Team pages)
• Recent news (search: "{company} funding 2024 2025")

Extract:

• What they build (products, services)
• Who they serve (customers, users)
• Recent news (funding, launches)
• Cultural signals

Gate 2 Check:

• Company name verified (exactly as in JD)
• Role title captured (exactly as in JD)
• 3+ company-specific details extracted

If Swiss company detected, load: references/swiss-market-rules.md

---

Step 4: Generate Content

Cover Letter

Structure (3-4 paragraphs, no headers):

Opening (50-70 words):

• Company-specific hook (NOT "I am writing to apply")
• Why this company interests you
• Your positioning

Body (200-280 words):

• 2-3 evidence paragraphs
• Each: Claim → Evidence → Result
• Include 2+ quantified achievements
• Include 3+ company-specific details

Closing (50-70 words):

• Confident (no "grateful")
• "I look forward to discussing..."

Total: 350-450 words

Style Rules (Non-Negotiable)

• NO em-dashes (use periods, commas instead)
• NO subjunctive ("would", "could", "might")
• Active voice 90%+
• Sentences: 10-20 words average
• Tone: Swiss formal, professional, neutral

Vocabulary

Never use: "passionate about", "excited to apply", "unique opportunity", "I would be grateful", "thrilled", "humbled"

Prefer: "I built", "I led", "resulting in", "measured by", "I look forward to"

For full rules, see references/writing-rules.md.

---

Step 5: Validate (Gate 3)

Self-check before presenting:

Content Checks

• Word count: 350-450
• Company-specific details: ≥3 (list them)
• Quantified achievements: ≥2 (list them)
• Company name triple-verified against JD
• Role title exact match to JD

Style Checks

• No blacklisted phrases
• No em-dashes (search for: —)
• No subjunctive mood (would, could, might, should)
• Active voice ≥90%
• Confident closing (not "grateful")

Metrics Verification

For each numeric claim:

1. Extract all metrics (patterns: X%, Xx, CHF X, X+)
2. Cross-reference against Metrics Bank in profile-essential.md
3. Mark VERIFIED or UNVERIFIED

If validation fails: Self-correct (max 2 iterations), then present with warning.

Quality Report (Collapsed)

Present to user:

Quality: PASSED ✓ (12/12 checks)
Word count: 412 | Company-specific: Yes | Metrics verified

[Show details]

---

Step 6: Create Output (Gate 4)

Create application folder with ALL files:

docs/applications/{company}/
├── job-description.md
├── cover-letter.md
├── cover-letter.html
├── cover-letter.pdf         ← MANDATORY
├── cv-{variant}.html        ← MANDATORY (copy from variants/)
├── cv-{variant}.pdf         ← MANDATORY
└── application-email.md

PDF Generation (MANDATORY - DO NOT SKIP)

Cover Letter PDF:

// Use Puppeteer to generate PDF
await page.pdf({
    path: 'cover-letter.pdf',
    format: 'A4',
    printBackground: true,
    margin: { top: 0, right: 0, bottom: 0, left: 0 }
});

CV Copy + PDF:

1. Copy CV variant from docs/profile/variants/cv-simon-yang-{variant}.html
2. Generate PDF with same Puppeteer settings
3. Both files go in application folder

Visual Verification (MANDATORY)

Take screenshot of cover-letter.html and verify:

• Fits on 1-2 pages max
• Professional formatting
• No broken layouts
• Print-ready appearance

Gate 4 Check:

• All placeholders replaced (no {{COMPANY}}, etc.)
• HTML renders correctly
• cover-letter.pdf generated (max 2 pages)
• CV variant copied to folder (HTML + PDF)
• Screenshot taken to verify print quality
• File structure complete (7 files minimum)

Cover letter HTML: Use template from docs/applications/_templates/cover-letter-template.html

Application email: Brief, Swiss formal, reference role and materials.

---

Metrics Bank (Quick Reference)

Use 2-3 metrics per cover letter. Full details in profile-essential.md.

Metric	Value	Use For
D30 Retention	37% vs 8%	Product, analytics
Retention Lift	125%	Data-driven decisions
Detection Rate	3x	Medical, preventive
Series A	CHF 11.5M	Executive, fundraising
Team Growth	1 to 8	Leadership
Solo Apps	5 production	Technical, GenAI

---

Reference Files

File	When Loaded	Tokens
references/job-types.md	Startup (@ import)	~800
references/profile-essential.md	After classification	~2,000
references/writing-rules.md	After clarification	~1,500
references/swiss-market-rules.md	If Swiss company	~600
references/quality-rules.md	During validation	~800

For full context loading order, see workflows/context-loading.md.

---

Example Invocation

User: /apply https://jobs.example.com/product-manager-ai

Claude: I'll analyze this job description.

[Scrapes JD]

I've classified this as **Type B: AI/ML Product Management**. Is that correct?

[Loads profile-essential.md]

What 2-3 aspects of this role interest you most?
Any specific achievement you want to emphasize?
Company tone: startup casual, corporate formal, or Swiss formal?
Any gaps to address?

[User responds]
[Loads writing-rules.md]

[Researches company, generates cover letter]

Here's your tailored cover letter:

[Cover letter text]

Quality: PASSED ✓ (12/12 checks)
Word count: 412 | Company-specific: Yes | Metrics verified

I recommend using **cv-simon-yang-ai-product.html** for this application.

Shall I create the application folder and save the files?

---

Success Criteria

1. Efficient context - Startup load <1,000 tokens
2. Correct classification - Job type matches JD keywords
3. Company research - Extracts 3+ specific details
4. Quality output - Passes all 12 validation checks
5. Human-like writing - No AI tells, reads naturally
6. Complete deliverables - ALL 7 files created:
   • job-description.md
   • cover-letter.md
   • cover-letter.html
   • cover-letter.pdf (MANDATORY)
   • cv-{variant}.html (MANDATORY)
   • cv-{variant}.pdf (MANDATORY)
   • application-email.md
7. Visual verification - Screenshot confirms print-ready quality